Image acquisition and analysis

Shape and alignment of excitation and depletion PSFs were checked at the beginning of every recording session to ensure optimal performance. After preparation, images of the cortical surface vascularization were taken under the preparation binocular for coarse reference. The mouse was then transferred to the microscope sample stage and the region of interest was identified in widefield mode at 10x magnification before switching to the 63x/1.30 objective. The tilt of the coverslip with respect to the focal plane was corrected with fine-pitch screws attached to the sample plate while observing the fluorescent beads under widefield illumination. As defocus of the beads was visi-ble to within±100nm over a field of view of 100µm , the coverslip tilt was effectively

reduced to±2 mrad (±0.1 deg.). Afterwards, the setup was switched to laser-scanning mode and target structures were identified performing fast overview scans. Vicinity to blood vessels and capillaries was avoided as hemoglobin slightly absorbs the or-ange STED light and occasionally caused thermal damage. The correction collar of the microscope objective was fine-tuned at the desired imaging depth by observing the image of a fine structure such as a thin axon while performing fast scans in the xz (or yz) plane. Imaging a sub-diffraction feature in this way effectively depicts the confocal PSF, which can then be optimized regarding its axial extent, focal plane symmetry and brightness. This procedure was repeated every time the imaging depth or the imaged region was changed.

For STED imaging, rectangular regions varying from 10x10µm to 30x30µm were raster-scanned with pixel sizes ranging from 20nm to 30nm at an 1:1 pixel aspect ratio.

As most structures of interest are not planar, up to 10µm thick volumes were imaged as stacks of xy-planes spaced 500-600nm apart. The time-average focal power of the excitation lasers was on the order of a 1-10 µW, while a maximum of 35mW was used for the STED laser. Where not indicated differently, images shown in this work were slightly smoothed with a Gaussian profile of 1.0 pixels width to suppress high-frequency noise. All quantitative analysis was performed on the raw data using ImageJ, including the depicted intensity profiles. For visualization of volume data, image stacks were projected to a single plane by maximum intensity projection. Image processing exceeding these steps is mentioned in the context of the respective measurements.

4.1 Dendritic spine plasticity in the adult brain

Dendritic spines are small protrusions that can be found on the dendrites of most prin-cipal neurons in the brain, such as on the pyramidal neurons of the neocortex. The dendrites of a single neuron can easily feature several hundred thousand of these pro-trusions. They form the post-synaptic part of the majority of excitatory synapses, with the pre-synaptic site mostly located on axons of other neurons. As Ramón y Cajal already depicted in his drawings (Fig. 1.1, Chap. 1), most spines have a pronounced head structure are connected to the dendrite by a thin neck. It is well known that den-dritic spines show a high degree of structural plasticity. They can change their shape on timescales of days to minutes and seconds including complete retraction of exist-ing spines andde-novogrowth of new ones [91]. Experiments have shown that these structural changes can be linked to synaptic activity¹ by correlating them to external stimuli [15,93–97]. Consequently, spines are generally attributed a key role in learn-ing processes and the formation of long-term memory. But even though experimental support for this hypothesis has recently started to emerge [98–100], the mechanisms

1Cajal was the first one to surmise that spine morphology and activity could be interdependent:"The state of activity would correspond then, to the swelling and elongation of the spines, and the resting state (sleep or inactivity) to their retraction."[92]

Fig. 4.1:Typical overview of a cortex region of a transgenic mouse prepared for imaging.

The left image is observed on a wide-field camera. The right picture is a maximum-intensity projection of a confocal overview stack extending over a depth of 5µm . Scale bar = 10µm .

Fig. 4.2:STED microscopy in the molecular layer of the somatosensory cortex of a mouse with EYFP labeled neurons. (A) Projected volumes of dendritic and axonal structures of an anesthetized mouse reveal (B) temporal dynamics of spine morphology with (C) a 4-fold improved resolution compared to diffraction-limited imaging. The curve is a three pixel wide line profile fitted to raw data with a Gaussian. Scale bars = 1µm . relating synaptic and structural plasticity are in large parts still poorly understood and therefore subject to intensive research

In typical experiments, the formation and elimination of entire spines is studied in liv-ing brain tissue over time with two-photon microscopy [101]. In addition to these rel-atively pronounced changes, however, there are also more subtle modifications, which affect the morphology of the spines on the nanoscale. As the spatial resolution of two-photon microscopes usually lies in the range of 250-400nm, these structural re-arrangements within individual spines are hard to quantify. Variations in the diameter of a spine neck, for example, lie in the range of 40 to 500nm [102]. The only way to currently study these small changes is to correlate the two-photon measurements with retrospective electron microscopy reconstructions [103,104]. Apart from being a time-consuming procedure, this does not allow the study of dynamic morphological features. It would therefore be advantageous if the detailed structure of single spines could be recorded by means of light- instead of electron microscopy. With STED mi-croscopy, movements of dendritic spines in hippocampal organotypical brain slices of newborn mice were observed with greatly enhanced spatial resolution [105]. Further-more, it was recently shown to be possible to resolve features of the underlying actin cytoskeleton, which is in large parts responsible the structural plasticity of the spines [19]. Until now it has, however, remained unclear whether such movements also oc-cur in the adult brain. Despite previous studies suggesting that spine motility is largely reduced during maturation [95], there are also hints that the remodeling of neuronal cir-cuits in the adult brain could involve such morphological rearrangements at the spine level [94].

The molecular layer of the somatosensory cortex of adult TgN(Thy1-EYFP) mice aged 2-12M was imagedin vivowith the upright STED microscope introduced earlier. The

Fig. 4.3:STED recordings of dendritic spines (A) reveal varying levels of spine motility among spines located in close proximity along the same dendrite (B, C). Scale bar = 1µm . mouse strain TgN(Thy1-EYFP) was chosen for the expression of the yellow fluores-cent protein (EYFP) in the cytoplasm of a subset of neurons [106,107]. Fig. 4.1 shows a typical overview of the cortex in wide-field and confocal mode after preparation.

Dendritic branches were chosen from the overview to record time-lapse series at high spatial resolution (Figs. 4.2, 4.3). The STED images show structure sizes of < 70 nm at the finest axons (Fig. 4.2C), indicating that the resolution is at least of that order.

Note-worthy are also the negatively contrasted, elongated structures along the den-drite, which could represent mitochondria displacing the stained cytosol (Fig. 4.2).

Time-lapse recordings on timescales of several minutes reveal that dendritic spines can indeed undergo continual morphologic changes and movements in the adult brain (Fig. 4.3B). To exclude random defocus from being mistaken for movement, each im-age was rendered by a maximum projection of a stack of 5 imim-ages with 600 nm depth spacing. While the dendrites of origin retained largely the same shape throughout the experiments, morphological changes were found at the head and neck regions of the dendritic spines, potentially reflecting alterations in the connectivity of the neural net-work. Motion was observed to be occurring at timescales ranging from a few seconds to minutes and to be of varying magnitude. Variations were thereby not only found be-tween different dendritic branches, but also bebe-tween closely spaced spines belonging to the same dendrite. A good example can be seen in Fig. 4.3. Even though morpholog-ical changes can be seen at the lower two spines of Fig. 4.3B, the filopodia-like spine at the top of the figure and the one depicted in 4.3C barely move during 15min. To get a better idea of the timescales involved, small stacks of single spines were recorded with STED at different rates. Fig. 4.4 shows a subset of a time series consisting of 128 projected stacks which were recorded at intervals of 10 seconds over a period of more than 20 minutes. Most changes at the spine head appear continuous at the depicted intervals of one minute, with few exceptions such as the small transient feature appear-ing only in the third frame (140s). Meanwhile, the small spine located closer to the dendrite does not show any movement.

Even though further quantification of the observed dynamics is beyond the scope of this thesis, it is worthwhile to discuss whether the movements could be induced by the new imaging modality itself. As no other microscopic technique is currently available to resolve features beyond the diffraction limit in the living brain, a definitive answer

Fig. 4.4:STED time-lapse recording of a single spine at a period of 10 seconds. The measure-ment includes 128 z-stacks consisting of 5 slices each. Showing only one of every six pictures in this overview, most of the rapid remodeling of the spine head still ap-pears continuous and smooth. No damage is observed at the dendrite or the spine after recording a total of 640 slices. Scale bar = 1µm .

cannot be given at this point. Coarse structural changes at larger spines, however, can readily be detected at confocal resolution. It turns out that timescale and type of the morphological changes observed with STED are not in contradiction with confocal measurements in brain slices reported earlier [108]. For direct comparison, the spine depicted in Fig. 4.4 was imaged every 60 seconds over a timespan of 15 minutes prior to switching on the STED beam (Appendix A.3). Here, also, similar motion can be ob-served, giving further evidence that STED microscopy does not induce motion beyond the levels seen in confocal microscopy.

Apart from possible functional impairments, interaction of the intense STED light with the tissue could also cause other effects. Although absorption of the 592 nm STED beam by EYFP is negligible, absorption by intrinsic chromophores in the tissue may occur at this wavelength. Indeed, slight local swellings have occasionally been ob-served for relatively thick dendritic processes featuring many mitochondria. We did not observe degradations or disaggregations of the processes, however, which have been shown to occur right after cell death or exitus. No evidence was found that STED microscopy introduces more stress to the neurons or to the surrounding tissue com-pared to established techniques as 2P-microscopy, where similar average focal power levels are commonly employed¹.